1.School of Science, Chang'an University, Xi'an 710064, China
2.Hangzhou DTWave Technology Co. , Ltd, Hangzhou 311100, China
Show less
文章历史+
Received
Accepted
Published
2023-09-15
2024-01-21
Issue Date
2025-10-09
PDF (810K)
摘要
属性约简是形式概念分析的重要研究问题,寻求高效的属性约简方法也是关注热点。本文在形式背景中引入矩阵粗糙熵,探讨对象导出三支概念格(Object-induced Three-way Concept Lattice,简称OE-概念格)上基于矩阵粗糙熵的属性约简方法。首先将OE-对象粒矩阵与粗糙熵相结合,定义OE-概念格的OE-矩阵粗糙熵,基于OE-对象粒矩阵的相似性提出OE-矩阵相似熵,研究它们的性质;其次在形式背景上定义OE-概念格的矩阵粗糙熵协调集和矩阵粗糙熵约简集,给出基于矩阵粗糙熵的协调集判定定理;利用OE-矩阵相似熵引入属性的重要性度量,进一步给出获取矩阵粗糙熵属性约简的方法和算法;最后在UCI(University of California, Irvine)数据集上进行属性约简实验,对于属性量小于10的数据集和属性量大于10的数据集,约简时间分别平均缩短了69.1%和97.5%,进而验证了该算法更适用于大样本数据。
Abstract
Attribute reduction is an important research issue in formal concept analysis, and seeking efficient attribute reduction methods is also a hot topic of concern. This paper studies attribute reductions of the object-induced three-way concept lattices by introducing matrix rough entropy into the formal context. Firstly, the OE-matrix (object-induced three-way matrix) rough entropy of the OE-concept lattice is defined by combining the OE-object granular matrix with rough entropy, and OE-matrix similarity entropy based on the similarity of OE-object granular matrix is proposed and the related properties of them are discussed. Secondly, the matrix rough entropy consistent set and matrix rough entropy reduction set of the OE-concept lattices are defined. Then the judgment theorems for the matrix rough entropy consistent set are given. The importance measure of attributes applying OE-matrix similarity entropy is proposed, by which the method and corresponding algorithm for obtaining matrix rough entropy attribute reduction are furtherly investigated. Finally, experiments of attribute reduction were conducted on the UCI (University of California, Irvine) dataset. The experimental results showed that for datasets with the count of attribute less than 10 and datasets with the count of attribute greater than 10, the reduction time was reduced by an average of 69.1% and 97.5%, respectively, confirming that the algorithm is more suitable for the large sample data.
由定理7和定理8可给出OE-概念格的矩阵粗糙熵约简算法:先求出每个属性的内重要度,得到核心属性集,进而判断核心属性集是否为矩阵粗糙熵约简,若不是,则通过计算外重要度给核心属性集中添加外重要度最大的属性,直到得到OE-概念格的矩阵粗糙熵协调集,从粗糙熵协调集中删除冗余属性得到约简。由此给出OE-概念格的矩阵粗糙熵约简算法(Algorithm of Attribute Reduction Based on Matrix-rough Entropy,ARMRE算法):
将表2中的UCI数据集转化为形式背景后进行属性约简实验,得到本文所提算法在各数据集下的约简时间,并与文献[29]中的算法进行约简时间的对比,最后通过绘制约简时间对比图,直观形象的总结出本文所提方法在约简时间上的优势。文献[29]受三支概念格粒约简的启发,基于OE-对象粒提出了基于信息熵的OE-概念格属性约简方法。本文结合OE-对象粒矩阵和粗糙熵提出了基于矩阵粗糙熵的OE-概念格属性约简方法,并用内重要度得到核心属性集,以外重要度对核心属性集增添属性来得到协调集,进而得到约简集的思想设计了算法。将文献[29]中的算法记为AROG(Algorithm of Attribute Reduction Based on Object-induced Granular)算法,用Matlab计算两种方法在以上十组数据集的约简时间,为了排除算法耗时随机性的影响,每个算法重复20次,然后计算其耗时平均值,对比两种算法的约简时间(表3)。
WILLER. Restructuring Lattice Theory: An Approach Based on Hierarchies of Concepts[C]//RIVAL I. Ordered Sets. Dordrecht: Springer, 1982: 445-470. DOI: 10.1007/978-94-009-7798-3_15 .
[2]
胡小康. 基于模糊关系的形式概念分析方法研究[D]. 太原: 山西大学, 2017.
[3]
HUX K. Research on Formal Concept Analysis Method Based on Fuzzy Relation[D]. Taiyuan: Shanxi University, 2017.
[4]
ZOUC F, DENGH F, WANJ F, et al. Mining and Updating Association Rules Based on Fuzzy Concept Lattice[J]. Future Gener Comput Syst, 2018, 82: 698-706. DOI: 10.1016/j.future.2017.11.018 .
[5]
HAOF, YANGY X, MING Y, et al. Incremental Construction of Three-way Concept Lattice for Knowledge Discovery in Social Networks[J]. Inf Sci, 2021, 578: 257-280. DOI: 10.1016/j.ins.2021.07.031 .
[6]
ZOUL, LINH M, SONGX Y, et al. Rule Extraction Based on Linguistic-valued Intuitionistic Fuzzy Layered Concept Lattice[J]. Int J Approx Reason, 2021, 133: 1-16. DOI: 10.1016/j.ijar.2020.12.018 .
[7]
QINK Y, LINH, JIANGY T. Local Attribute Reductions of Formal Contexts[J]. Int J Mach Learn Cybern, 2020, 11(1): 81-93. DOI: 10.1007/s13042-019-00956-z .
ZHANGW X, WEIL, QIJ J. Theory and Method of Attribute Reduction of Concept Lattice[J]. Sci China Ser E, 2005, 35(6): 628-639. DOI: 10.1360/112004-104 .
[10]
QIJ J. Attribute Reduction in Formal Contexts Based Onanewdiscernibility Matrix[J]. J Appl Math Comput, 2009, 30(1): 305-314. DOI: 10.1007/s12190-008-0174-9 .
WANQ, MAY C, LIJ H. Three-way Concept Acquisition and Attribute Characteristic Analysis Based on Pictorial Diagrams[J]. J Front Comput Sci Technol, 2022, 16(12): 2879-2889. DOI: 10.3778/j.issn.1673-9418.2104120 .
WANR X, LIM, MIAOD Q. Characterization of Formal Concept Analysis for Complete Lattice in the Clarified Attribute Dual Context[J]. J Shanxi Univ Nat Sci Ed, 2022, 45(1): 77-86. DOI: 10.13451/j.sxu.ns.2021063 .
[17]
CORNEJOM E, MEDINAJ, RAMÍREZ-POUSSAE. On the Use of Irreducible Elements for Reducing Multi-adjoint Concept Lattices[J]. Knowl Based Syst, 2015, 89: 192-202. DOI: 10.1016/j.knosys.2015.07.003 .
[18]
WANGZ, QIJ J, SHIC J, et al. Multiview Granular Data Analytics Based on Three-way Concept Analysis[J]. Appl Intell, 2023, 53(11): 14645-14667. DOI: 10.1007/s10489-022-04145-4 .
[19]
NIUJ J, CHEND G. Incremental Calculation Approaches for Granular Reduct in Formal Context with Attribute Updating[J]. Int J Mach Learn Cybern, 2022, 13(9): 2763-2784. DOI: 10.1007/s13042-022-01561-3 .
[20]
WANGZ, SHIC J, WEIL, et al. Tri-granularity Attribute Reduction of Three-way Concept Lattices[J]. Knowl Based Syst, 2023, 276: 110762. DOI: 10.1016/j.knosys.2023.110762 .
[21]
QIJ J, WEIL, YAOY Y. Three-way Formal Concept Analysis[C]//International Conference on Rough Sets and Knowledge Technology. Cham: Springer, 2014: DOI: 732-741.10.1007/978-3-319-11740-9_67 .
[22]
YAOY Y. An Outline of a Theory of Three-way Decisions[C]. Berlin Heidelberg: Springer, 2012: 1-17. DOI: 10.1007/978-3-642-32115-3_1 .
[23]
RENR S, WEIL. The Attribute Reductions of Three-way Concept Lattices[J]. Knowl Based Syst, 2016, 99: 92-102. DOI: 10.1016/j.knosys.2016.01.045 .
ZHANGC L, LIJ J, LINY D. Attribute Reduction in Formal Contexts Based on OE-concept Lattices[J]. Comput Eng Appl, 2021, 57(15): 82-89. DOI: 10.3778/j.issn.1002-8331.2006-0325 .
[28]
SHANNONC E. A Mathematical Theory of Communication[J]. Bell Syst Tech J, 1948, 27(3): 379-423. DOI: 10.1002/j.1538-7305.1948.tb01338.x .
[29]
LIJ L, HEZ Y, ZHUQ L. An Entropy-based Weighted Concept Lattice for Merging Multi-source Geo-ontologies[J]. Entropy, 2013, 15(12): 2303-2318. DOI: 10.3390/e15062303 .
[30]
SINGHP K, CHERUKURIA K, LIJ H. Concepts Reduction in Formal Concept Analysis with Fuzzy Setting Using Shannon Entropy[J]. Int J Mach Learn Cybern, 2017, 8(1): 179-189. DOI: 10.1007/s13042-014-0313-6 .
[31]
LIANGJ Y, XUZ B. The Algorithm on Knowledge Reduction in Incomplete Information Systems[J]. Int J Unc Fuzz Knowl Based Syst, 2002, 10(1): 95-103. DOI: 10.1142/s021848850200134x .
HUANGB, ZHOUX Z, SHIY C. Entropy of Knowledge and Rough Set Based on General Binary Relation[J]. Syst Eng Theory Pract, 2004, 24(1): 93-96. DOI: 10.3321/j.issn: 1000-6788.2004.01.016 .
HEQ Q, MAJ M, DINGN. Matrix Information Entropy Based Matrix Entropy Reduction in Formal Contexts[J]. J Nanjing Univ Nat Sci, 2023, 59(1): 98-106. DOI: 10.13232/j.cnki.jnju.2023.01.010 .